engineermike
Well-Known Member
- Thread starter
- #1
I started down the path of transmission tuning using PCMTec and, man, there's a lot here. I have had success transplanting shift characters and even making hybrid shift characters (normal shifts at part throttle and drag shifts at full throttle, for example), but I might just give up understanding how the tuning logic actually works. There are something like 15 parameters that control shift properties, not including shift schedule, torque modulation, or shift pressures (significantly more parameters). I was hoping to figure out how they all work together to control a shift. Parameter names are like oncoming ramp (time), offgoing ramp (delta time), torque rate offgoing, torque transfer offgoing, torque rate oncoming, inertia phase with 5 sub-phases (time), and many more.
Couple things I figured out:
- There are no pressure sensors in the transmission. All of the pressures are assumed based on the electrical power sent to the solenoids.
- There are 3 speed sensors in the trans. The logic calculates the slip of all the clutches based on these 3 speeds.
- There are 6 clutches and combinations of them create each gear. For every upshift, 4 clutches stay in their same state, one releases, and another applies simultaneously. For the clutches that remain applied before during and after the shift, the pressure is increased, presumably to prevent them from slipping during the shock of the shift.
- There is line pressure, and there is solenoid pressure. The line pressure is supplied to the solenoids, then the solenoids modulate it from there. So adding line pressure doesn't necessarily increase pressure applied to the clutch.
For the sake of logistics and logging, I decided to study the 3-4 shift. In that shift, the E clutch releases and the F clutch is applied. As I often preach, you have to reduce the channel list to a bare minimum if you want a high enough data rate to log things that happen quickly.
This is what a part-throttle 3-4 shift looks like:
The cursor is placed at the point in time where the E clutch starts releasing and the F clutch starts grabbing. What's initially striking to me about this is that the clutches don't actually start swapping until a solid .7 seconds after the F solenoid sends pressure, while the actual shift only took about .3 seconds to complete. I guess this explains why part throttle paddle shifts are so laggy. Also noteworthy is that the line pressure is increased to 1150 kpa during the entirety of the shift. 1150 coincides with the "Trans Line Pressure Power On Upshift" table in HPT. Also, the initial pressure spike from the F solenoid appears to align with the "Trans boost pressure element F" table. Boost pressure is the pressure intended to just fill the cavities with fluid before the clutches are applied. You can also see that the torque modulation takes place just during the clutch swap time period. What I can't quite grasp is a) the fact that nothing in the way of solenoid actuation seems to trigger the switch from clutch E to F and b) how the dozens of torque rate, torque transfer, ramp time, inertia time, and other parameters play into this specifically.
The following is a similar snip of a higher torque shift:
On this shift, the actual shift only took about .1 seconds to complete, but the solenoids started moving about .4 seconds before the shift. The boost pressure and line pressure do not seem to follow the tables identified above that appeared to control it. Torque modulation appears to have started earlier, and once again no real defining event on the solenoid side indicating it should switch gears at that moment; just the pressure balance I guess.
Another thing I wondered about is what governs the clutch pressure between shifts.
In this 3rd gear steady part of the log, it appears as though line pressure and clutch pressure modulate with torque, but there is also a minimum "floor" they aren't dropping below. However, the floor for the E clutch pressure of 313 kpa and line pressure of 410 kpa don't seem to appear in the tune file so I don't know where they are sourced from. To make matters more complicated, we know there is some "adaptive" learning that takes place, which can only happen if there are PID feedback loops. However, the only input to the transmission is solenoid electrical power and the only measurable output is clutch slip, which would imply the PID feedback loop is to relate clutch slip to solenoid power. This adds another layer of complexity.
So basically, after all that, I only have observations and definitely no explanation as to what all of the torque rates, torque transfers, ramp times, etc. do and how they interact to control the slip and shift firmness.
If anyone knows anything about how the shifts are controlled, I'd be interested in hearing it.
Couple things I figured out:
- There are no pressure sensors in the transmission. All of the pressures are assumed based on the electrical power sent to the solenoids.
- There are 3 speed sensors in the trans. The logic calculates the slip of all the clutches based on these 3 speeds.
- There are 6 clutches and combinations of them create each gear. For every upshift, 4 clutches stay in their same state, one releases, and another applies simultaneously. For the clutches that remain applied before during and after the shift, the pressure is increased, presumably to prevent them from slipping during the shock of the shift.
- There is line pressure, and there is solenoid pressure. The line pressure is supplied to the solenoids, then the solenoids modulate it from there. So adding line pressure doesn't necessarily increase pressure applied to the clutch.
For the sake of logistics and logging, I decided to study the 3-4 shift. In that shift, the E clutch releases and the F clutch is applied. As I often preach, you have to reduce the channel list to a bare minimum if you want a high enough data rate to log things that happen quickly.
This is what a part-throttle 3-4 shift looks like:
The cursor is placed at the point in time where the E clutch starts releasing and the F clutch starts grabbing. What's initially striking to me about this is that the clutches don't actually start swapping until a solid .7 seconds after the F solenoid sends pressure, while the actual shift only took about .3 seconds to complete. I guess this explains why part throttle paddle shifts are so laggy. Also noteworthy is that the line pressure is increased to 1150 kpa during the entirety of the shift. 1150 coincides with the "Trans Line Pressure Power On Upshift" table in HPT. Also, the initial pressure spike from the F solenoid appears to align with the "Trans boost pressure element F" table. Boost pressure is the pressure intended to just fill the cavities with fluid before the clutches are applied. You can also see that the torque modulation takes place just during the clutch swap time period. What I can't quite grasp is a) the fact that nothing in the way of solenoid actuation seems to trigger the switch from clutch E to F and b) how the dozens of torque rate, torque transfer, ramp time, inertia time, and other parameters play into this specifically.
The following is a similar snip of a higher torque shift:
On this shift, the actual shift only took about .1 seconds to complete, but the solenoids started moving about .4 seconds before the shift. The boost pressure and line pressure do not seem to follow the tables identified above that appeared to control it. Torque modulation appears to have started earlier, and once again no real defining event on the solenoid side indicating it should switch gears at that moment; just the pressure balance I guess.
Another thing I wondered about is what governs the clutch pressure between shifts.
In this 3rd gear steady part of the log, it appears as though line pressure and clutch pressure modulate with torque, but there is also a minimum "floor" they aren't dropping below. However, the floor for the E clutch pressure of 313 kpa and line pressure of 410 kpa don't seem to appear in the tune file so I don't know where they are sourced from. To make matters more complicated, we know there is some "adaptive" learning that takes place, which can only happen if there are PID feedback loops. However, the only input to the transmission is solenoid electrical power and the only measurable output is clutch slip, which would imply the PID feedback loop is to relate clutch slip to solenoid power. This adds another layer of complexity.
So basically, after all that, I only have observations and definitely no explanation as to what all of the torque rates, torque transfers, ramp times, etc. do and how they interact to control the slip and shift firmness.
If anyone knows anything about how the shifts are controlled, I'd be interested in hearing it.
Sponsored
